A digital service supply chain collaborative processing method

Abstract

The application relates to the technical field of supply chain analysis, in particular to a digital service supply chain collaborative processing method, which comprises the following steps: a demand fluctuation is analyzed through a trend analysis algorithm in a central processing unit to generate a demand fluctuation characteristic model; based on the demand fluctuation characteristic model, fault tolerance thresholds of each node are set, the fault tolerance thresholds comprising a resource redundancy range, a logistics delay tolerance and an inventory buffer amount; on the basis of the demand fluctuation characteristic model and the fault tolerance thresholds, AI algorithms are used to predict future demand peaks and valleys of the supply chain in real time, and the resource allocation of each node is dynamically adjusted through resource allocation; a collaborative fault tolerance mechanism among nodes is established, so that when a node exceeds the fault tolerance threshold, the demand pressure is relieved by adjusting the resources of other nodes. The application ensures that the system can flexibly respond and quickly adjust under the condition of large demand fluctuation, thereby improving the adaptability and reaction speed of the supply chain.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of supply chain analysis, and particularly relates to a digital service supply chain collaborative processing method. BACKGROUND

[0002] Modern supply chains usually involve multiple nodes, including suppliers, manufacturers, warehousing facilities, distribution centers, and retailers, each of which has its unique needs and resource allocation. However, factors such as demand fluctuations, resource shortages, and logistics delays often lead to instability and inefficiency in supply chain operations, especially in market environments with large demand fluctuations. Ensuring coordination and stability among various links of the supply chain has become a technical problem that needs to be solved urgently.

[0003] In traditional supply chain management, many enterprises often rely on historical data to predict demand and allocate resources accordingly. However, demand prediction often has a high degree of uncertainty, especially when market demand fluctuates greatly or unexpected events occur. Traditional demand prediction methods cannot respond to sharp fluctuations in demand in real time, leading to problems such as overstocking or shortages, waste of transportation resources, and affecting the overall efficiency of the supply chain and customer satisfaction.

[0004] Currently, the focus is mainly on improving the accuracy of demand prediction models and optimizing resource management at a single node. The coordination and fault tolerance mechanism between nodes in the supply chain is relatively weak. In particular, when faced with complex and dynamic demand fluctuations, how to achieve cross-node collaborative allocation through intelligent scheduling remains a technical bottleneck in the field of supply chain management. SUMMARY

[0005] The present application provides a digital service supply chain collaborative processing method.

[0006] A digital service supply chain collaborative processing method includes the following steps:

[0007] S1, demand fluctuation data collection and trend analysis: real-time monitoring of demand changes at each node of the supply chain, including order volume, inventory consumption rate, and logistics demand rate, through a data collection module, and aggregating the data to a central processing unit. The central processing unit analyzes demand fluctuations through a trend analysis algorithm to generate a demand fluctuation feature model.

[0008] S2, dynamic setting of fault tolerance threshold: based on the demand fluctuation feature model, set the fault tolerance threshold of each node, including the resource redundancy range, the logistics delay tolerance, and the inventory buffer amount, to ensure that the supply chain can flexibly allocate resources under demand fluctuations, while avoiding resource waste;

[0009] S3, flexible resource allocation and adaptive optimization: based on the demand fluctuation characteristic model and fault tolerance threshold, the AI algorithm is used to predict the future demand peak and valley of the supply chain in real time, and the resource allocation of each node is dynamically adjusted, including increasing or reducing inventory, allocating standby transportation resources, to ensure that each node meets the demand during the demand peak period and avoids excessive redundancy during the demand trough period;

[0010] S4, inter-node collaborative fault tolerance mechanism: an inter-node collaborative fault tolerance mechanism is established, so that when a node exceeds the fault tolerance threshold, the demand pressure is relieved by allocating resources of other nodes, and the inter-node collaborative fault tolerance is monitored and scheduled in real time by the central processing unit, including cross-node inventory sharing, logistics resource redistribution and flexible scheduling of human resources.

[0011] Optionally, the data acquisition module in S1 specifically comprises:

[0012] The order quantity data is obtained in real time through the order management system of each node, and the generation, cancellation and change information of the order is directly synchronized to the data acquisition module to provide the latest order quantity change information, and the order quantity is represented as O t , wherein t is the current time;

[0013] The inventory data is extracted from the inventory management system of each node, and the inventory consumption rate R t is calculated, and the inventory consumption rate is generated based on the change of the initial inventory and the real-time outbound data;

[0014] The logistics demand data is obtained through the logistics scheduling system, including the real-time state of the transportation task, the planned traffic volume and the actual shipment quantity information, and the information is summarized to the data acquisition module to ensure the timeliness and accuracy of the logistics demand data, and then the logistics demand rate Q t is calculated.

[0015] Optionally, the demand fluctuation is analyzed by a trend analysis algorithm in the central processing unit in S1 to generate a demand fluctuation characteristic model, specifically comprising:

[0016] S11, demand fluctuation characteristic extraction:

[0017] S111, fluctuation amplitude: the standard deviation of the order quantity, the inventory consumption rate and the logistics demand rate is calculated to measure the fluctuation amplitude, and the order quantity fluctuation amplitude characteristic Amplitude O , the inventory consumption rate fluctuation amplitude characteristic Amplitude R , and the logistics demand rate fluctuation amplitude characteristic Amplitude Q are obtained.

[0018] S112, periodicity feature: using wavelet transform to analyze the periodicity of order quantity, inventory consumption rate and logistics demand rate, identify the periodicity of demand fluctuation, including order quantity periodicity feature Period O , inventory consumption rate periodicity feature Period R , and logistics demand rate periodicity feature Period Q ;

[0019] S113, trend feature: using moving average method to extract the long-term trend of data, the trend feature of order quantity is T O , the trend feature of inventory consumption rate is T R , and the trend feature of logistics demand rate is T Q ;

[0020] S12, construction of demand fluctuation feature model: based on the above extracted fluctuation amplitude, periodicity and trend features, construct a demand fluctuation feature model M d , the demand fluctuation feature model M d includes the key features of supply chain demand in the fluctuation process, and the demand fluctuation feature model M d is expressed as:

[0021] M d =

[0022] {Amplitude O ,Amplitude R ,Amplitude Q ,Period O ,Period R ,Period Q ,T O ,T R ,T Q}。

[0023] Optionally, the setting of the resource redundancy range comprises: according to the order quantity fluctuation amplitude feature Amplitude O and the periodicity feature Period R of inventory consumption rate, and setting the resource redundancy range to ensure that each node can provide sufficient redundant resources to meet the demand during the fluctuation amplitude or periodic demand peak; according to the order quantity fluctuation amplitude, a redundancy coefficient is set to measure the strength of demand fluctuation and the amount of redundant resources, and the size of the amount of redundant resources should be able to meet the scheduling capability of each node of the supply chain during the surge of order quantity, through the analysis of the periodicity feature of inventory consumption rate, the peak and trough periods of inventory consumption are identified, and the redundant resource configuration is increased during the periodic peak to avoid supply interruption due to insufficient inventory during the demand peak.

[0024] Optionally, the setting of the logistics delay tolerance includes: setting the logistics delay tolerance based on the periodic characteristic Period of the logistics demand rate Q and the trend characteristic T of the logistics demand rate Q , the logistics delay tolerance reflects the tolerance range of the supply chain to transportation delay, providing buffer space for each node during the surge of logistics demand, and reducing the impact of logistics delay on the operation of the supply chain; the logistics delay tolerance needs to consider the fluctuation of logistics demand during the period of large demand fluctuation (such as promotion season or holiday), and during the period of large demand fluctuation, the logistics demand surges, and a sufficient time window should be reserved to cope with transportation delay or other time bottlenecks in the supply chain, and the tolerance is adjusted according to the intensity of demand fluctuation: during the period of demand surge, the delay tolerance is increased to avoid supply chain interruption caused by logistics delay.

[0025] Optionally, the setting of the inventory buffer amount includes: setting the inventory buffer amount according to the fluctuation amplitude Amplitude of the inventory consumption rate R and the long-term trend characteristic T of the inventory consumption rate R , the inventory buffer amount ensures that the node inventory is sufficient during the demand surge to avoid inventory depletion or shortage, and ensures the smooth operation of the supply chain; the fluctuation amplitude of the inventory consumption rate reflects the change range of the inventory consumption, and the long-term trend characteristic of the inventory consumption rate indicates the overall change direction of the inventory consumption, the purpose of setting the inventory buffer amount is to ensure that each node has sufficient inventory to meet the demand during the demand surge, and avoid the occurrence of supply interruption or inventory shortage;

[0026] During the demand peak period (such as order quantity surge or consumption increase), there should be sufficient buffer inventory to cope with sudden demand. While in the demand trough period (such as seasonal low demand period), inventory redundancy should be reduced to avoid inventory accumulation.

[0027] Optionally, the S3 includes establishing a demand prediction model and training, through the collection of historical data, the demand prediction model, combined with order quantity, inventory consumption rate and logistics demand factors, using neural network to predict the future demand fluctuation of the supply chain, including the peak and valley of demand, after the demand prediction model is trained, the demand data of each node is collected in real time and input into the prediction model, the demand prediction in the future period is carried out, and the peak and valley of demand in the future short period is predicted.

[0028] Optionally, the input variables of the demand prediction model based on neural network include O t , R t , Q t , and the output variables include the predicted demand for future time step t+k (1-3 days), focusing on the peak and valley of demand;

[0029] The structure of the neural network is divided into an input layer, a hidden layer and an output layer, wherein:

[0030] The input layer includes order quantity, inventory consumption rate and logistics demand rate at time step t as input features;

[0031] The hidden layer includes multiple neurons, which are nonlinearly mapped using a ReLU activation function;

[0032] The output layer outputs a predicted demand quantity The linear activation function is used to predict the demand peak and the demand valley.

[0033] Optionally, after the demand prediction model is trained, the demand peak and the demand valley are extracted:

[0034] Demand peak prediction: in the prediction result, the time point with the maximum predicted value is selected as the demand peak: representing the maximum demand in the future demand prediction period; Demand valley prediction: in the prediction result, the time point with the minimum predicted value is selected as the demand valley:

[0035] representing the minimum demand in the future demand prediction period. Optionally, the S4 specifically includes:

[0036] S41, fault tolerance threshold monitoring and identification: real-time monitoring of the demand state of each node, including order quantity, inventory consumption rate and logistics demand, when the demand of a node exceeds the set fault tolerance threshold (such as exceeding the set resource redundancy range or logistics delay tolerance), the central processing unit will automatically identify that the node enters a high load state, and start the cooperative fault tolerance mechanism;

[0037] S42, resource allocation decision generation: the central processing unit generates a resource allocation decision based on global demand fluctuation data and resource conditions between nodes, and the decision includes:

[0038] Inventory sharing: when the inventory of a node is lower than the preset fault tolerance threshold, the required inventory is allocated from other nodes with sufficient inventory through a cross-node inventory sharing mechanism;

[0039] Logistics resource redistribution: when the logistics demand of a node exceeds the tolerance range, the idle transportation resources of other nodes are allocated to the node with high demand pressure through a cross-node logistics resource redistribution mechanism.

[0040] The present application has the following beneficial effects:

[0041] The present application has the following beneficial effects:

[0042] ​The present application, by establishing a resource allocation mechanism based on demand fluctuation characteristic model and AI prediction, realizes the dynamic response of the supply chain when facing demand fluctuation, especially in the peak and trough period of demand, can adjust the resource configuration of each node in real time, avoids the problem of excess or shortage of inventory, in this way, the overall operation efficiency of the supply chain is greatly improved, and the resource utilization is optimized, ensures that in the case of large demand fluctuation, the system can flexibly respond and quickly adjust, thereby improving the adaptability and reaction speed of the supply chain.

[0043] The present application, the inter-node cooperative fault tolerance mechanism, through the real-time monitoring and scheduling of the central processing unit, when a certain node exceeds the fault threshold, the resources of other nodes can be automatically allocated to relieve the demand pressure. This mechanism covers cross-node inventory sharing, logistics resource redistribution and human resource scheduling, significantly enhances the flexibility and stability of the supply chain, combines resource allocation with demand fluctuation prediction, so that the system can quickly fill the resource gap through the cooperation of other nodes when the node pressure is too large, ensures the stable operation of the supply chain, and avoids the negative impact of single node overload on the overall supply chain. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only a part of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0045] Fig. 1 The flowchart of the cooperative processing method of the embodiment of the present application is shown in the figure.

[0046] Fig. 2 The inter-node cooperative fault tolerance mechanism of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0047] The present application will be described in detail below in combination with the drawings and specific embodiments. It should be noted here that in order to make the embodiments more detailed, the following embodiments are the best, preferred embodiments, and other alternative ways can also be used by those skilled in the art to implement; and the drawings are only used to describe the embodiments more specifically, and are not intended to specifically limit the present application.

[0048] It is noted that the use of "one embodiment," "an embodiment," "certain embodiments," "some embodiments," etc., in the specification does not necessarily refer to the same embodiment, although it may. In addition, the description is not intended to limit the scope of the application to the described or illustrated embodiments. Further, the use of the term "based on" in the description is intended to be given its ordinary and customary meaning to an ordinary skilled person in the field of the application, and is not intended to foreclose implementations that can be understood as being based on, or in addition to, indications other than the one(s) now described.

[0049] In general, the terminology or nomenclature used can be understood at least partially from the context in which the terminology or nomenclature is used. For example, depending at least partially on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular or can be used to describe combinations of features, structures, or characteristics that are combinable into one or more instances of a feature, structure, or characteristic. In addition, the term "based on" can be understood as not necessarily of requiring a complete set of factors, but, at least partially, allowing a lesser set of factors to come within the scope of the application.

[0050] As shown in FIG. 1, a method for collaborative processing based on a digitized service supply chain includes the following steps: Figs. 1-2

[0051] S1, demand fluctuation data acquisition and trend analysis: real-time monitoring of demand changes of each node in the supply chain through a data acquisition module, including order quantity, inventory consumption rate and logistics demand rate, and the data is aggregated to a central processing unit, and the central processing unit analyzes the demand fluctuation through a trend analysis algorithm to generate a demand fluctuation characteristic model;

[0052] S2, dynamic setting of fault tolerance threshold: setting the fault tolerance threshold of each node based on the demand fluctuation characteristic model, including the resource redundancy range, the logistics delay tolerance and the inventory buffer quantity, to ensure that the supply chain can flexibly allocate resources under demand fluctuation, while avoiding resource waste;

[0053] S3, flexible resource allocation and adaptive optimization: based on the demand fluctuation characteristic model and the fault tolerance threshold, using AI algorithm to predict the future demand peak and valley of the supply chain, dynamically adjusting the resource configuration of each node through resource allocation, including increasing or reducing inventory, allocating standby transportation resources, to ensure that each node meets demand during demand peak period and avoids excessive redundancy during demand trough period;

[0054] S4, inter-node collaborative fault tolerance mechanism: establishing an inter-node collaborative fault tolerance mechanism, so that when a node exceeds the fault tolerance threshold, the demand pressure is relieved by allocating resources of other nodes, and the inter-node collaborative fault tolerance is monitored and scheduled in real time by the central processing unit, including cross-node inventory sharing, logistics resource redistribution and flexible scheduling of human resources.

[0055] ​The data collection module in S1 specifically includes:

[0056] The order quantity data is obtained in real time through the order management system of each node, and the generation, cancellation and change information of the order is directly synchronized to the data collection module to provide the latest order quantity change information, and the order quantity is represented as O t , where t is the current time;

[0057] The inventory data is extracted from the inventory management system of each node, and the inventory consumption rate R t is calculated, which is generated based on the change of the initial inventory and the real-time outbound data to ensure the accurate update of the inventory status, and the inventory consumption rate R t is represented as: Where S0 is the initial inventory, S t is the real-time inventory at the current time, and represents the percentage reduction of the initial inventory;

[0058] The logistics demand data is obtained through the logistics scheduling system, including the real-time state of the transportation task, the planned logistics flow and the actual shipment quantity information, which is summarized to the data collection module to ensure the timeliness and accuracy of the logistics demand data, and then the logistics demand rate Q t is calculated, which is expressed as:

[0059] Where D plan is the current planned logistics demand quantity (total quantity of planned shipment), and D actual is the actual shipment quantity, and the formula reflects the difference between the planned demand and the actual shipment, and when the logistics demand satisfaction rate is lower than the threshold, the data collection module can issue an alarm prompt to adjust the logistics resource allocation in time to avoid the supply chain disruption.

[0060] The demand fluctuation is analyzed by the trend analysis algorithm in the central processing unit in S1 to generate a demand fluctuation feature model, which specifically includes:

[0061] S11, demand fluctuation feature extraction:

[0062] S111, fluctuation amplitude: the fluctuation amplitude is measured by calculating the standard deviation of the order quantity, the inventory consumption rate and the logistics demand rate to obtain the order quantity fluctuation amplitude feature Amplitude O , the inventory consumption rate fluctuation amplitude feature Amplitude R , and the logistics demand rate fluctuation amplitude feature Amplitude Q , for example, the fluctuation amplitude of the order quantity can be represented as: Where O i is the order quantity at each time, For the average value of order quantity, similarly, the fluctuation range of inventory consumption rate and logistics demand rate can be calculated;

[0063] S112, periodicity feature: using wavelet transform to analyze the periodicity of order quantity, inventory consumption rate and logistics demand rate, identify the periodicity of demand fluctuation, including order quantity periodicity feature Period O , inventory consumption rate periodicity feature Period R , and logistics demand rate periodicity feature Period Q ;

[0064] Taking order quantity periodicity analysis as an example, for example, using wavelet transform to analyze the periodicity of order quantity, through wavelet transform, different frequency components in order quantity data can be identified, and then significant periodicity rules can be found, assuming that order quantity data is O t (t = 1, 2,..., N), the specific steps of wavelet transform are as follows:

[0065] 1. Select mother wavelet: adopt Morlet wavelet function, which is defined as:

[0066] Where ω0is the center frequency of the wavelet;

[0067] 2. Calculate wavelet coefficients: for each scale s and position t, the wavelet transform coefficient W(s, t) can be obtained by convolution of order quantity data O t and mother wavelet function:

[0068] Where s controls the scale (frequency component) of the wavelet, larger s value corresponds to low frequency component, and smaller s value corresponds to high frequency component.

[0069] 3. Identify periodicity: by drawing wavelet energy diagram or wavelet power spectrum, observe the energy density under different scales s, if the energy is concentrated at a certain scale (i.e. frequency), the data corresponding to the scale has significant periodicity, assuming that the scale corresponding to the energy concentration area is s * , then the corresponding period P is about: P = s * · Δt, where Δt is the time step, the period P obtained by wavelet transform is the order quantity periodicity feature Period O .

[0070] Similarly, the periodicity features of inventory consumption rate and logistics demand rate can be calculated.

[0071] S113, trend feature: using moving average method to extract the long-term trend of data, the trend feature of order quantity is T O , the trend feature of inventory consumption rate is T R, the trend characteristics of the logistics demand rate is T Q ;

[0072] Taking order volume trend analysis as an example, the long-term trend of order volume is extracted using the moving average method. Through moving average processing, short-term fluctuations in order volume are eliminated, and long-term trend characteristics are retained. Order volume data is O t (t = 1, 2,..., N), and its long-term trend T t is obtained as follows:

[0073] 1. Select window size: Set a window size w, and select a larger w value (one week or one month) to filter out short-term fluctuations and retain only long-term trends;

[0074] 2. Calculate the moving average: For each time t, use the moving average formula with window size w to calculate the long-term trend T t : The formula represents the average value of order volume in the window [t-w+1, t], which is the long-term trend of the current time t. The T t sequence obtained by moving average processing is the long-term trend characteristics T O of order volume. Moving average can effectively remove short-term fluctuations, making the long-term growth or decline trend of order volume clearer, thus more accurately reflecting the demand trend of the supply chain.

[0075] Similarly, the trend characteristics of inventory consumption rate and logistics demand rate can be calculated.

[0076] S12, construction of demand fluctuation feature model: based on the extracted fluctuation amplitude, periodicity and trend characteristics, construct the demand fluctuation feature model M d , the demand fluctuation feature model M d includes the key features of supply chain demand in the fluctuation process, and the demand fluctuation feature model M d is expressed as:

[0077] M d =

[0078] {Amplitude O , Amplitude R , Amplitude Q , Period O , Period R , Period Q , T O , T R , T Q}.

[0079] The setting of resource redundancy range includes: according to the order volume fluctuation amplitude characteristics AmplitudeO and the periodicity of inventory consumption rate Period R and set the resource redundancy range to ensure that each node can provide sufficient redundant resources to meet the demand when the fluctuation range is large or the demand peak period is periodic; according to the fluctuation range of order quantity, a redundancy coefficient is set to measure the strength of demand fluctuation and the amount of redundant resources, and the amount of redundant resources should be sufficient to meet the scheduling capability of each node in the supply chain when the order quantity surges, therefore, the setting of the amount of redundant resources needs to reflect the range of demand fluctuation and ensure the continuity of the supply chain, so as to avoid resource shortage due to sudden increase of orders, in addition, through the analysis of the periodicity of inventory consumption rate, the peak and trough periods of inventory consumption are identified, and the allocation of redundant resources is increased during the periodic peak period to avoid supply interruption due to insufficient inventory during the demand peak period.

[0080] The setting of logistics delay tolerance includes: based on the periodicity of logistics demand rate Period Q and the trend of logistics demand rate T Q The delay tolerance range is preset for the logistics demand peak period, and the logistics delay tolerance reflects the tolerance range of the supply chain to transportation delay, so as to provide buffer space for each node when the logistics demand surges, and reduce the impact of logistics delay on the operation of the supply chain; the logistics delay tolerance needs to consider the fluctuation of logistics demand during the period of large demand fluctuation (such as promotion season or holiday), and sufficient time window should be reserved to cope with transportation delay or other time bottlenecks in the supply chain during the period of large demand fluctuation, and the tolerance is adjusted according to the strength of demand fluctuation: the delay tolerance is increased during the period of demand surge, so as to avoid the interruption of the supply chain caused by logistics delay.

[0081] By this method, the logistics tolerance not only considers the short-term demand fluctuation, but also combines the long-term logistics demand trend, so that the system can maintain a certain flexibility in the changing environment and ensure the stable operation of the supply chain.

[0082] The setting of inventory buffer amount includes: according to the fluctuation amplitude of inventory consumption rate Amplitude R and the long-term trend of inventory consumption rate T R The inventory buffer amount of each node is set, which ensures that the node inventory is sufficient when the demand surges, so as to avoid the occurrence of inventory depletion or shortage and ensure the smooth operation of the supply chain; the fluctuation amplitude of inventory consumption rate reflects the change range of inventory consumption amount, and the long-term trend of inventory consumption rate indicates the overall change direction of inventory consumption, the purpose of setting the inventory buffer amount is to ensure that each node has sufficient inventory to meet the demand when the demand surges, so as to avoid the occurrence of supply interruption or inventory shortage;

[0083] During peak demand periods (e.g. when order volume surges or consumption increases), there should be sufficient buffer inventory to cope with sudden demand. While in the trough of demand (e.g. seasonal low demand period), inventory redundancy should be reduced to avoid inventory accumulation;

[0084] In addition, by analyzing the long-term trend of inventory consumption, we can identify the long-term change pattern of inventory demand, such as seasonal demand change or long-term growth trend, and adjust the configuration of inventory buffer accordingly to ensure that each node of the supply chain can still meet the changing inventory demand in the future demand changes.

[0085] S3 includes the establishment of demand prediction model and training, through the collection of historical data, demand prediction model, combined with order volume, inventory consumption rate and logistics demand factors, using neural network to predict the future demand fluctuation of supply chain, including the peak and trough of demand, after the training of demand prediction model, real-time collection of demand data of each node, and input into the prediction model, to predict the demand in the future period, and predict the demand peak and trough in the future short term.

[0086] The input variables of the demand prediction model based on neural network include O t , R t , Q t , and the output variables include The predicted demand for future time step t+k (1-3 days) is concerned about the peak and trough of demand;

[0087] The structure of neural network is divided into input layer, hidden layer and output layer, in which:

[0088] Input layer: including order volume, inventory consumption rate and logistics demand rate at time step t as input features;

[0089] Hidden layer: multiple neurons, using ReLU activation function for nonlinear mapping;

[0090] Output layer: output predicted demand Use linear activation function to predict demand peak and trough.

[0091] The training process includes the following steps:

[0092] 1. Data preparation: collect historical data, including order volume O t , inventory consumption rate R t , logistics demand rate Q t , and standardize or normalize the data to ensure that the data will not be affected by the model performance due to the large scale difference during training.

[0093] The training set and validation set are divided, using 80% of the time series data as the training set and the remaining 20% as the validation set.

[0094] 2. Model training: The neural network is trained using the backpropagation method, aiming to minimize the error between the predicted value and the true demand O t+k The loss function is the mean square error (MSE): where, Oi is the actual demand of the i-th sample, O is the demand predicted by the neural network.

[0095] 3. Adam optimizer to adjust the network weights to minimize the loss function.

[0096] 4. Tuning during training: Adjust the number of layers of the network, the number of neurons in each layer, the learning rate, and other hyperparameters to optimize the model.

[0097] After the demand prediction model is trained, extract the demand peak and demand valley:

[0098] Demand peak prediction: In the prediction results, select the time point with the maximum predicted value as the demand peak: representing the maximum demand in the future demand prediction period;

[0099] Demand valley prediction: In the prediction results, select the time point with the minimum predicted value as the demand valley: representing the minimum demand in the future demand prediction period.

[0100] Once the peak and valley of demand are predicted, dynamic adjustment of resource allocation at each node can be based on these prediction results:

[0101] At the demand peak:

[0102] Increase inventory: Increase the inventory according to the predicted demand peak to ensure that demand is met.

[0103] Allocate standby transportation resources: Increase the allocation of transportation resources according to the demand peak prediction to ensure timely delivery during high demand periods.

[0104] At the demand valley:

[0105] Reduce inventory: Reduce the inventory to avoid overstocking and reduce resource waste.

[0106] Optimize transportation resources: Reduce the allocation of standby transportation resources to avoid idle transportation resources.

[0107] S4 specifically includes:

[0108] S41, fault tolerance threshold monitoring and identification: real-time monitoring of the demand state of each node, including order volume, inventory consumption rate and logistics demand, when the demand of a node exceeds the set fault tolerance threshold (such as exceeding the set resource redundancy range or logistics delay tolerance), the central processing unit will automatically identify that the node enters a high load state, and start the cooperative fault tolerance mechanism;

[0109] S42, resource allocation decision generation: the central processing unit generates resource allocation decisions based on global demand fluctuation data and resource conditions between nodes, including:

[0110] Inventory sharing: when the inventory of a node is lower than the preset fault tolerance threshold, the required inventory is allocated from other nodes with sufficient inventory through cross-node inventory sharing mechanism;

[0111] Logistics resource redistribution: when the logistics demand of a node exceeds the tolerance range, idle transportation resources of other nodes are allocated to the node with high demand pressure through cross-node logistics resource redistribution mechanism;

[0112] The central processing unit monitors the state changes of each node in real time, including order processing, inventory consumption rate, logistics resource usage and human resource allocation, and executes resource allocation instructions in real time through dynamic scheduling, to ensure that cross-node resources can be flexibly allocated when the demand of the node fluctuates, and the overall stability of the supply chain is ensured.

[0113] The present application covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits are not described in detail.

[0114] The above is only the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application.

Claims

1. A method for collaborative processing based on digitized service supply chain, characterized in that, Comprise the following steps: S1, demand fluctuation data acquisition and trend analysis: through the data acquisition module real-time monitoring of the supply chain nodes of demand change, including order quantity, inventory consumption rate and logistics demand rate, the data to the central processing unit, central processing unit through trend analysis algorithm to analyze the demand fluctuation, generate demand fluctuation characteristic model; S2, fault tolerance threshold dynamic setting: based on the demand fluctuation characteristic model, set the fault tolerance threshold of each node, the fault tolerance threshold includes resource redundancy range, logistics delay tolerance and inventory buffer; S3, resource flexible deployment and adaptive optimization: on the basis of demand fluctuation characteristic model and fault tolerance threshold, using AI algorithm to predict the future demand peak and trough of supply chain, through resource deployment dynamic adjustment of resource configuration of each node, including increasing or reducing inventory, distribution of standby transportation resources, to ensure that each node meets the demand in demand peak period, avoid too much redundancy in demand trough; S4, inter-node collaborative fault tolerance mechanism: establish inter-node collaborative fault tolerance mechanism, when a node exceeds the fault tolerance threshold, through the deployment of other node resources to relieve demand pressure, the inter-node collaborative fault tolerance is monitored and dispatched by the central processing unit, including cross-node inventory sharing, logistics resource redistribution and flexible scheduling of human resources; The central processing unit in S1 analyzes the demand fluctuation through the trend analysis algorithm to generate the demand fluctuation characteristic model, specifically comprising: S11, demand fluctuation feature extraction: S111, fluctuation amplitude: the fluctuation amplitude is measured by calculating the standard deviation of the order quantity, inventory consumption rate and logistics demand rate, and the order quantity fluctuation amplitude feature Amplitude is obtained O , the inventory consumption rate fluctuation amplitude feature Amplitude R , the logistics demand rate fluctuation amplitude feature Amplitude Q ; S112, periodicity feature: using wavelet transform to analyze the periodicity of order quantity, inventory consumption rate and logistics demand rate, identify the periodicity of demand fluctuation, including order quantity periodicity feature Period O , inventory consumption rate periodicity feature Period R , logistics demand rate periodicity feature Period Q ; S113, trend feature: long-term trend of data is extracted by using moving average method, the trend feature of order quantity is T O , the trend feature of inventory consumption rate is T R , and the trend feature of logistics demand rate is T Q ; S12, construction of the demand fluctuation characteristic model: based on the extracted fluctuation amplitude, periodicity and trend characteristics, the demand fluctuation characteristic model M is constructed d , the demand fluctuation characteristic model M d includes the key characteristics of the supply chain demand in the fluctuation process, the demand fluctuation characteristic model M d is expressed as: M d ＝ {Amplitude O ,Amplitude R ,Amplitude Q ,Period O ,Period R ,Period Q ,T O ,T R ,T Q} The setting of the resource redundancy range comprises: according to the order quantity fluctuation amplitude characteristic Amplitude O and the periodicity characteristic Period of inventory consumption rate R , and setting the resource redundancy range to ensure that each node can provide sufficient redundant resources to meet the demand during the period of large fluctuation amplitude or high peak demand; according to the order quantity fluctuation amplitude, a redundancy coefficient is set to measure the strength of demand fluctuation and the amount of redundant resources, and the size of the amount of redundant resources should be able to meet the scheduling capability of each node of the supply chain during the surge of order quantity; through the analysis of the periodicity characteristic of inventory consumption rate, the peak period and the trough period of inventory consumption are identified, and the redundant resource configuration is increased during the periodic peak period to avoid the supply interruption due to the lack of inventory during the demand peak period. The setting of the logistics delay tolerance includes: periodically setting the logistics delay tolerance based on the periodic characteristics Period of the logistics demand rate Q and the trend characteristics T of the logistics demand rate Q The delay tolerance range is preset for the peak period of logistics demand, and the logistics delay tolerance reflects the tolerance range of the supply chain to transportation delay. The logistics delay tolerance needs to consider the fluctuation of logistics demand in the period with large demand fluctuation. In the period with large demand fluctuation, the logistics demand increases sharply, and sufficient time window should be reserved to cope with transportation delay or other time bottlenecks in the supply chain. The tolerance is adjusted according to the intensity of demand fluctuation: in the period with increased demand, the delay tolerance is increased to avoid supply chain interruption caused by logistics delay. The setting of the inventory buffer quantity comprises: setting the inventory buffer quantity of each node according to an inventory consumption rate fluctuation amplitude characteristic Amplitude R and an inventory consumption rate long-term trend characteristic T R , the inventory buffer quantity ensuring that the node inventory is sufficient when the demand surges; the inventory consumption rate fluctuation amplitude characteristic reflecting the change range of the inventory consumption quantity, the inventory consumption rate long-term trend characteristic indicating the overall change direction of the inventory consumption, the purpose of setting the inventory buffer quantity being to ensure that each node has sufficient inventory to meet the demand when the demand surges, avoiding the situation of supply interruption or inventory shortage; During the demand peak period, there should be enough buffer inventory to cope with sudden demand, while in the demand trough, inventory redundancy is reduced to avoid inventory accumulation.

2. The method according to claim 1, wherein, The data acquisition module in S1 specifically comprises: The order quantity data is acquired in real time through the order management system of each node, and the generation, cancellation and change information of the order is directly synchronized to the data acquisition module to provide the latest order quantity change information, and the order quantity is represented as O t where t is the current time. extracting inventory data from inventory management systems of each node and calculating an inventory consumption rate R t , the inventory consumption rate being generated based on changes in inventory initial amount and real-time outbound data; The logistics demand data is acquired through the logistics scheduling system, including the real-time state of the transportation task, the planned logistics flow and the actual shipping volume information, and then the logistics demand rate Q is calculated t .

3. The method according to claim 2, wherein, The S3 includes establishing a demand prediction model and training, through the collection of historical data, demand prediction model, combined with order quantity, inventory consumption rate and logistics demand factors, using neural network to predict the future demand fluctuation of supply chain, including demand peak and trough, after the demand prediction model training is completed, real-time acquisition of demand data of each node, and input into the prediction model, to predict the demand in the future period, predict the demand peak and trough in the short term in the future.

4. The method according to claim 3, wherein, The demand prediction model based on neural networks input variables include O t , R t , Q t , and output variables include forecast demand for future time step t+k, focusing on peaks and valleys of demand; The structure of neural network is divided into input layer, hidden layer and output layer, wherein: Input layer: including order quantity, inventory consumption rate and logistics demand rate of time step t as input features; Hidden layer: multiple neurons, using ReLU activation function for nonlinear mapping; Output layer: Outputs predicted demand amounts The demand peaks and troughs are predicted with a linear activation function.

5. The method according to claim 4, wherein, After the demand prediction model training is completed, the demand peak and demand trough are extracted: Demand peak prediction: in the prediction result, select the predicted value The maximum time point as the demand peak: Represent the maximum demand that occurs within the prediction period of future demand; Demand valley prediction: in the prediction result, select the predicted value The minimum time point as the demand valley: Represents the minimum demand that occurs within the prediction period of future demand.

6. The method of claim 1, wherein the method further comprises: The S4 specifically comprises: S41, fault tolerance threshold monitoring and identification: real-time monitoring of the demand state of each node, including order quantity, inventory consumption rate and logistics demand, when the demand of a node exceeds the set fault tolerance threshold, the central processing unit will automatically identify that the node enters high load state, and start the collaborative fault tolerance mechanism; S42, resource deployment decision generation: the central processing unit generates resource deployment decision based on global demand fluctuation data and inter-node resource status, the decision includes: Inventory sharing: when the inventory of a certain node is lower than the preset fault tolerance threshold, the required inventory is allocated from other nodes with sufficient inventory through cross-node inventory sharing mechanism; Logistics resource redistribution: when the logistics demand of a certain node exceeds the tolerance range, the idle transportation resources of other nodes are allocated to the node with high demand pressure through cross-node logistics resource redistribution mechanism.

Images